Independent component analysis minimizing convex divergence

Authors:
Yasuo Matsuyama;Naoto Katsumata;Ryo Kawamura
Affiliations:
Department of Computer Science, Waseda University, Tokyo, Japan;Department of Computer Science, Waseda University, Tokyo, Japan;Department of Computer Science, Waseda University, Tokyo, Japan
Venue:
ICANN/ICONIP'03 Proceedings of the 2003 joint international conference on Artificial neural networks and neural information processing
Year:
2003

Citing 7
Cited 1

Blind separation of sources, Part 1: an adaptive algorithm based on neuromimetic architecture

Signal Processing
An information-maximization approach to blind separation and blind deconvolution

Neural Computation
Adaptive online learning algorithms for blind separation: maximum entropy and minimum mutual information

Neural Computation
Natural gradient works efficiently in learning

Neural Computation
The Alpha-EM Algorithm: A Block Connectable Generalized Learning Tool for Neural Networks

IWANN '97 Proceedings of the International Work-Conference on Artificial and Natural Neural Networks: Biological and Artificial Computation: From Neuroscience to Technology
Equivariant adaptive source separation

IEEE Transactions on Signal Processing
The α-EM algorithm: surrogate likelihood maximization using α-logarithmic information measures

IEEE Transactions on Information Theory

Promoter Recognition for E. coli DNA Segments by Independent Component Analysis

CSB '04 Proceedings of the 2004 IEEE Computational Systems Bioinformatics Conference

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new class of learning algorithms for independent component analysis (ICA) is presented. Starting from theoretical discussions on convex divergence, this information measure is minimized to derive new ICA algorithms. Since the convex divergence includes logarithmic information measures as special cases, the presented method comprises faster algorithms than existing logarithmic ones. Another important feature of this paper's ICA algorithm is to accept supervisory information. This ability is utilized to reduce the permutation indeterminacy which is inherent in usual ICA. By this method, the most important activation pattern can be found as the top one. The total algorithm is tested through applications to brain map distillation from functional MRI data. The derived algorithm is faster than logarithmic ones with little additional memory requirement, and can find task related brain maps successfully via conventional personal computer.